How to Block AI from Training on Your Images
AI companies scrape billions of images from the web to train their models. without asking permission. The ai:opt-out=true metadata standard gives you a way to signal that your work is off-limits. This guide explains how it works and how to embed it into your files in under a minute.
Why AI crawlers are scraping your photos
Every time you publish a photo online. on your portfolio, Instagram, a stock site, or even a client delivery. it becomes visible to web crawlers. AI companies run large-scale crawling operations to collect training data for image generation and recognition models.
Unlike traditional search engine crawlers, AI training crawlers don't index your page for search results. they download your images to use as raw training material. The result: your original photography, illustration, or artwork ends up inside a model that can reproduce your style without attribution or payment.
The robots.txt file on your website can block specific crawlers from your domain. But that only works if the image stays on your domain. Once a file is downloaded, emailed, re-uploaded, or shared, the website-level block no longer applies. Metadata embedded inside the file travels with it.
The key insight: robots.txt protects your website. ai:opt-out metadata protects the file itself. wherever it ends up.
What is the ai:opt-out XMP tag?
XMP (Extensible Metadata Platform) is an Adobe standard for embedding structured metadata inside image and document files. It's already used for EXIF data, copyright information, GPS coordinates, and camera settings. The ai:opt-out key is an extension of this standard specifically for AI training signals.
When ai:opt-out=true is embedded in a file's XMP packet, it tells crawlers: the creator of this file does not consent to its use in AI training datasets.
The signal lives in the binary XMP header segment of the file. It survives:
- File downloads and re-uploads
- Sharing via email or messaging apps
- Hosting on third-party platforms (where they preserve metadata)
- Being embedded in documents or presentations
Which AI crawlers honour the opt-out tag?
The following major AI companies publicly commit to respecting ai:opt-out signals in image metadata:
This does not mean every AI company in existence will honour it. Smaller or less scrupulous scrapers may ignore metadata entirely. But the largest, most commercially significant AI training operations. the ones whose models people actually use. do respect the signal.
How to embed the ai:opt-out tag in 3 steps
VeriMedia embeds the tag directly in your browser. No uploads, no accounts, no software to install.
-
Open VeriMedia Go to verimedia.xyz. The tool runs entirely in your browser. Chrome, Firefox, Safari, or Edge all work.
-
Drop your file onto the tool Drag and drop your JPEG, PNG, WebP, HEIC, or PDF onto the upload area, or click to browse. Files up to 25 MB are supported on the free tier. Your file is read into browser memory. it never leaves your device.
-
Download the shielded file VeriMedia embeds
ai:opt-out=truein the XMP packet, strips GPS coordinates and device fingerprints, and returns the clean file for download. The whole process takes under 5 seconds.
Tip for photographers: Run your files through VeriMedia as the last step before any upload. before posting to your portfolio, sending to clients, or publishing on social media. This ensures every file that leaves your machine carries the opt-out signal.
What else gets embedded or removed?
VeriMedia does more than the opt-out tag in a single pass:
- GPS coordinates removed. your precise location at the time of shooting is stripped from the file
- Device fingerprints removed. camera make, model, serial number, and lens data are cleared
- AI opt-out embedded.
ai:opt-out=truewritten to the binary XMP packet - Creator metadata added (Plus/Pro). your name, copyright statement, and portfolio URL embedded in standard XMP fields
- Google Licensable Badge (Pro). the
WebStatementXMP tag that triggers a "Licensable" badge in Google Images
Does it affect image quality?
No. XMP metadata is stored in a dedicated header segment inside the file, completely separate from the pixel data. Embedding or removing metadata does not change the visual content of your image in any way. File size may decrease slightly (typically 5–50 KB) because the old metadata block is replaced with a leaner one.
Frequently asked questions
Can I use this on already-published photos?
Yes. you can process and re-upload them. However, copies already downloaded by crawlers before you embedded the tag may have been processed. The tag is most effective when embedded before first publication.
Does this work on iPhone photos (HEIC)?
Yes. VeriMedia handles HEIC files natively. They are converted to a standard format with the opt-out tag embedded and returned as a downloadable file.
Is the ai:opt-out tag a legal protection?
It is a policy signal, not a legal instrument. Companies that publicly commit to honouring it are making a voluntary pledge. For stronger legal protection, pair file-level metadata with clear copyright notices and licensing terms on your website.
What's the difference between ai:opt-out and robots.txt?
robots.txt tells crawlers not to visit specific URLs on your website. It only works while the image lives on your domain. ai:opt-out is embedded in the file itself and travels with it. so it still signals opt-out even after the image is downloaded, shared, or re-hosted.
Does VeriMedia upload my files?
No. Every step. reading the file, embedding metadata, stripping GPS. happens inside your browser using JavaScript. No bytes are sent to any server at any point.
Shield your images now. free
Embed ai:opt-out=true and strip private metadata in under 5 seconds. No account needed.